apertium-deu-fin: Finnish–German rules for rule-based machine translation

This is a visualisation of some rules in apertium transfer.

Categories (parts of chunks)

These are the categories Apertium is using in order to chunk, re-order and transfer lexemes.

Category	Items
noun	<n.*> <np.*>
adj	<adj.*>
num	<num.*>
negverb	<vaux.neg.*> <vblex.neg.*>
verb	<vblex.*> <vaux.*> <vbmod.*> <vbser.*>
someprn	<prn.rel.*> <prn.dem.*> <det.def.*>
inf	<vblex.*.infa.*> <vaux.*.infa.*> <vbmod.*.infa.*> <vbser.*.infa.*> <vblex.*.infma.*> <vaux.*.infma.*> <vbmod.*.infma.*> <vbser.*.infma.*>
pastverb	<vblex.*.past.*> <vaux.*.past.*> <vbmod.*.past.*> <vbser.*.past.*>
sent	<sent>

Attributes

These are the morphological analysis value (tag) sets that can be processed in the transfer.

Attribute set name	Tags
a_case	<nom> <ine> <ela> <ill> <ade> <abl> <all> <par> <gen> <acc> <tra> <lat> <ess> <dat>
a_noun	<n> <np> <n.abbr> <n.acr> <np.abbr> <np.acr>
a_prn	<prn> <prn.dem> <prn.rel> <det.def>
a_adj	<adj>
a_num	<num>
a_gender	<m> <f> <nt> <mf>
a_number	<sg> <pl> <sp>
a_verb	<vblex> <vblex.neg> <vblex.sep> <vaux> <vbser> <vbmod>
a_voice	<actv> <pasv>
a_tense	<pri> <pii> <ifi> <pres> <past>
a_prsnum	<p1.sg> <p2.sg> <p3.sg> <p1.pl> <p2.pl> <p3.pl>

Macros

Macros are helper functions in apertium transfer files.

test

Parametres: 1

1. let $number ≔ “”

num-mangler

Parametres: 1

1. if sl[1]['lem'] ≟ “yksi” then:
2. let tl[1]['a_number'] ≔ <sg>
3. else:
4. let tl[1]['a_number'] ≔ <pl>

case-mangler

Parametres: 1

1. if sl[1]['a_case'] ≟ <ela>sl[1]['a_case'] ≟ <ill>sl[1]['a_case'] ≟ <abl>sl[1]['a_case'] ≟ <all>sl[1]['a_case'] ≟ <par>sl[1]['a_case'] ≟ <acc> then:
2. let tl[1]['a_case'] ≔ <acc>
3. elseif sl[1]['a_case'] ≟ <ine>sl[1]['a_case'] ≟ <ade> then:
4. let tl[1]['a_case'] ≔ <dat>
5. elseif sl[1]['a_case'] ≟ <ins>sl[1]['a_case'] ≟ <tra>sl[1]['a_case'] ≟ <ess> then:
6. let tl[1]['a_case'] ≔ <nom>

tensemood-mangler

Parametres: 1

1. if sl[1]['a_tense'] ≟ <past> then:
2. let tl[1]['a_tense'] ≔ <pii>
3. elseif sl[1]['a_tense'] ≟ <pri> then:
4. let tl[1]['a_tense'] ≔ <pri>

Rules

The actual rules concerning stuff.

negverb neg: find persnum and fix

Matching pattern:

1. negverb
2. verb

Action:

1. tensemood-mangler($1)
2. Output:
3. vp<VP> 1. tl[2]['lem'] tl[2]['a_verb'] tl[2]['a_tense'] tl[1]['a_prsnum'] 1. blank1 1. “nicht”<adv>

Compose syntactic past form

Matching pattern:

1. pastverb

Action:

1. tensemood-mangler($1)
2. Output:
3. vp<VP.PAST> 1. “haben”<vbhaver.pri>tl[1]['a_prsnum'] 1. blank0 1. tl[1]['lem'] tl[1]['a_verb'] <pp>

Map infs sto inf

Matching pattern:

1. inf

Action:

1. tensemood-mangler($1)
2. Output:
3. vp<VP> 1. tl[1]['lem'] tl[1]['a_verb'] <inf>

Drop voice from verbs, mangle tense mood

Matching pattern:

1. verb

Action:

1. tensemood-mangler($1)
2. Output:
3. vp<VP> 1. tl[1]['lem'] tl[1]['a_verb'] tl[1]['a_tense'] tl[1]['a_prsnum']

Mangle case of some prns but not all

Matching pattern:

1. someprn

Action:

1. case-mangler($1)
2. Output:
3. np<NP> 1. tl[1]['lem'] tl[1]['a_prn'] tl[1]['a_gender'] tl[1]['a_number'] tl[1]['a_case']

Mangle case of nouns

Matching pattern:

1. noun

Action:

1. case-mangler($1)
2. Output:
3. np<NP> 1. tl[1]['lem'] tl[1]['a_noun'] tl[1]['a_gender'] tl[1]['a_number'] tl[1]['a_case']

Mangle case of adjs

Matching pattern:

1. adj

Action:

1. case-mangler($1)
2. Output:
3. np<AP> 1. tl[1]['lem'] tl[1]['a_adj'] <sint.attr>

Drop case, mangle num of nums… German is weirdly tagged

Matching pattern:

1. num

Action:

1. num-mangler($1)
2. Output:
3. nump<NumP> 1. tl[1]['lem'] tl[1]['a_num'] tl[1]['a_number']

Default rule

Matching pattern:

1. sent

Action:

1. Output:
2. sent<SENT> 1. tl[1]['whole']

---

Documentation for apertium-deu-fin. Generated with Flammie’s apevis-xslt.